Lecture Notes in Computer Science, 2001, Volume 2159/2001, 539-542, DOI: 10.1007/3-540-44811-X_61

Emergence of Cooperative Tactics by Soccer Agents with Ability of Prediction and Learning

Yoichiro Kumada and Kazuhiro Ueda

View Related Documents

Abstract

This paper describes soccer agents that can learn by themselves to make use of emergent cooperative tactics. Considering the methods that actual coaches of soccer help their players to learn the way of good play, we developed a mechanism of learning to distinguish good tactics from bad ones. The agents adaptively learned the utilities of the situations and the conditional probabilities about the situations in order to predict the opponents’ behavior and in order to select good actions. The agents cooperatively created some combinations of passes, such as a wall pass and a one-two pass which we had never taught in advance how to execute.

Fulltext Preview

Image of the first page of the fulltext document